The Effect of Branch Banking on Financial Stability

Branch Banking, Bank Competition, and
Financial Stability
Mark Carlson
and
Kris James Mitchener*
February 2003
Version 1.3
JEL Codes: G21, N22, E44
It is often argued that branching stabilizes banking systems by facilitating diversification of bank
portfolios; however, previous empirical research on the Great Depression cannot be reconciled
with this view. Analyses using state-level data find that states allowing branch banking had lower
failure rates, while those examining individual banks find that branch banks were more likely to
fail. We argue that an alternative hypothesis can reconcile these seemingly disparate findings.
Using data on national banks from the 1920s and 1930s, we show that branch banking increases
competition and forces weak banks to exit the banking system. This consolidation strengthens the
system as a whole without necessarily strengthening the branch banks themselves.
*
Carlson: Federal Reserve Board; Mail Stop 86; 20th and Constitution Ave, NW; Washington DC 20551;
[email protected]. Mitchener: Department of Economics, Santa Clara University and NBER; 500
El Camino Real; Santa Clara, CA 95053; [email protected]. The authors thank Waiyi Poon for valuable
research assistance and Joe Mason, Bill Sundstrom, David Wheelock, and participants at the Allied Social
Sciences Meetings, Cliometric Society Sessions for comments and suggestions. The views presented in this
paper are solely those of the authors and do not necessarily represent those of the Federal Reserve System
or its staff.
Branch Banking, Bank Competition, and Financial Stability
One of the foundations of the theoretical literature on banking regulation is that
branch banking leads to more stable banking systems by enabling banks to better
diversify their assets and widen their depositor base (Gart 1994, Hubbard 1994). This
conventional wisdom has been used to argue that historical banking crises in the United
States, especially those of the 1930s, would have been less severe had the U.S. permitted
widespread branch banking (Friedman and Schwartz 1963, Calomiris 2000). The
empirical literature examining banking stability in the Great Depression, however, has
not universally confirmed this prediction. In fact, this research presents a paradox.
Studies using aggregate bank failure data from the Depression find that states that
allowed branch banking had lower failure rates than those that only allowed unit banking
(Wheelock 1995, Mitchener 2000a, 2000b), while studies using individual bank data find
that banks that had branches were more likely to fail than unit (or single office) banks
(Calomiris and Mason 2000, Carlson 2001).
In this paper, we resolve this empirical puzzle by focusing on an alternative
channel through which branch banking affects financial stability: increased competition.
Our hypothesis is that, faced with heightened competition, banks that are only marginally
profitable are forced out of the banking system either through merger or voluntary
liquidation. As these weaker banks close, the overall stability of a state’s banking system
improves through consolidation.
This explanation allows us to reconcile the paradox of the Great Depression, as
there could be lower failure rates in states allowing branch banking without the branch
banks themselves being the strongest banks. This new view suggests that at the onset of
1
the Great Depression, there were still many weak banks in states prohibiting branch
banking; the real shock of the 1930s caused many of these to fail. However, in states that
permitted branching, weak banks had been pruned from the system, and failures were
consequently lower at the systemwide level. Our hypothesis also fills a void in the
broader literature on financial stability. According to Allen and Gale (2000), the
relationship between stability and competition has received little previous attention.1
Our hypothesis has several testable propositions, including one that enables us to
discriminate our view from the diversification hypothesis. First, if branching does remove
weaker banks from the system, then states that permit branch banking should experience
higher merger and voluntary liquidation rates and lower entry rates than states prohibiting
branching. Second, if the lower failure rate in branch banking states are the result of weak
banks exiting the system rather than diversification, then failure rates should be more
strongly related to a measure of prior consolidation within the banking industry than an
indicator of branch banking.
Since our hypothesis emphasizes changes in the competitive environment of
banking induced by the onset of branching, it is necessary to test our model using data
from a period when branch banking was relatively new and expanding in scope.
Moreover, because we want to test how branching influences the stability of banking
systems, we also need to examine a period when there were numerous failures. In this
respect, the experience of the U.S. banking system from 1920-30 is ideal since branching
was in its infancy, but expanding rapidly (Figure 1), and because the 1920s were
1
“On the one hand, there are many models of competition in the literature including models of bank
regulation in a competitive environment. On the other hand, there is a well-developed literature on bank
crises…But there is little on the impact of competition on stability.” (Allen and Gale, 2000, p.268).
2
characterized by a large number of bank failures (Figure 2).2 Moreover, studying this
period allows us to compare our results to existing research, which has used data from the
1920s and 1930s to draw conclusions about the effects of branching.
Our empirical results support the predictions of the competition hypothesis. States
that adopted branching laws experienced more mergers and voluntary liquidations and
less entry during the 1920s. We also find that, although there was significant
consolidation in the banking sector in states allowing branch banking, profits were lower
on average in these states, suggesting that branching led to increased competition rather
than monopoly power. To test whether branching reduced failures, we first confirm that
our data produce the usual state level result that states allowing branches statewide had
lower failure rates. Then we include an index of prior consolidation and find, as predicted
by the competition hypothesis, that this measure outperforms the branching indicator.
Thus, industry consolidation and the removal of weak and inefficient unit banks from the
system are shown to be a better predictor of failure rates than measures of branch
banking.3
The paper proceeds as follows. Section 2 discusses the previous literature on
branching and financial stability. In section 3, we present our hypothesis for resolving the
existing puzzle in the literature. The next section tests the consolidation hypothesis and
some of its implications for bank competition and financial stability. Section 5 provides
concluding remarks.
2
Chapman and Westerfield (1942) describe how the issue of branch banking gained national attention
during the 1920s, in part because it was spreading rapidly in states such as California and prompting federal
regulators to reconsider their longstanding prohibitions against it.
3
We also compare the results of our hypothesis with an alternative: that branching increases the ability of
banks to merge rather than fail by providing a larger pool of prospective merger partners. We thank Joe
Mason for suggesting this alternative.
3
Section 2. The Effects of Branch Banking on Financial Stability
An argument commonly articulated in the literature is that branch banking
stabilizes banking systems by reducing their vulnerability to local economic shocks:
branching enables banks to diversify their loans and deposits over a wider geographical
area or customer base.4 Restrictions on branching have been linked to the instability of
banking systems. Calomiris (2000) argues that bank failures were more prevalent in
regions of the United States without branch banking as well as in countries lacking it.
Friedman and Schwartz (1963) suggest that the absence of branching in the U.S.
increased the severity of the banking panics during the Great Depression. Moreover, they
argue that the U.S. experience stands in contrast to Canada, which experienced banking
distress during the Depression but not widespread failures and a collapse of its banking
system.5 The view that branch banking stabilizes banking systems through diversification
is in fact an old argument. In the 1920s, it was used by proponents of branch banking to
encourage state legislators to adopt laws legalizing branch banking (Preston, 1924 and
Southworth, 1928).
Studies examining the effects of branching at the aggregate level generally
support the hypothesis that allowing branch banking increases systemic stability.
Wheelock (1995) studies the relationship of different regulations on bank failure rates in
different states during the years 1929-1932 and finds that states that allowed branch
4
Studies by Wacht (1968) and Lauch and Murphy (1970) find reduced variance in deposit flows for branch
banks. Cherin and Melcher (1988) find that branching has moderating effects on asset returns.
4
banking tended to have lower failure rates. Mitchener (2000a, 2000b) further examines
state- and county-level bank failure rates. Controlling for economic fundamentals and
differences in both state supervision and regulation, he also finds that states with
legalized branching had lower failure rates during the Depression. Comparing 25
different countries during the Great Depression, Grossman (1994) finds that those with
large branching networks were less likely to experience banking crises. And using data
from the postwar period in the U.S., Rose and Scott (1978) find that bank failures from
1946-75 were concentrated in states limiting or prohibiting branch banking.6 Although
these studies (using aggregate data) find a positive correlation between branch banking
and financial stability, they do not establish a direct link between the purported benefits
of diversification of loan portfolios and the deposit base to financial stability.7
Studies using data on individual banks operating in the 1920s and 1930s paint a
different picture of the effects that branching has on the survivorship of individual banks,
and they cast doubt on the common view that the stabilizing benefits of branching operate
via increased diversification opportunities. Calomiris and Mason (2000) find that, during
the Depression, Federal Reserve members that were branch banks tended to fail sooner
than unit banks. Also using data on individual state banks from this period, Carlson
5
Drummond (1991) and White (1983) make a similar argument. Kryzanowski and Roberts (1993),
however, find that nationwide branch banking did not prevent banks in Canada from becoming technically
insolvent.
6
Many of these studies also include some measure of bank concentration as an explanatory variable.
Although none of them interacts branch banking and their measure of concentration, it is possible that these
two effects worked together to influence failure rates. In section 4, we relate our consolidation hypothesis
to the issue of banking concentration. Our emphasis, however, differs in that we are testing whether
changes in bank concentration are an outcome of branching laws, and whether this in turn affected bank
profitability.
7
Wheelock (1995) attributes the positive correlation between restrictions on branch banking and failure
rates to limited diversification. Similarly Alston, Grove, and Wheelock (1994) also consider the impact of
branching legislation on bank failures, and emphasize that branching may reduce a bank’s susceptibility to
distress in a particular area; however, they do not find that the ratio of (non-home-office) branches to total
banks helps to explain the cross-state variation in failure rates during the 1920s.
5
(2001) examines three states where branch banking was relatively widespread and finds
that branched banks were more likely to fail than unit banks. Furthermore, he rejects
some potential reasons for this phenomenon including insufficient diversification and
over-expansion on the part of banks. Instead, he finds that branch banks used
diversification to reduce their reserves rather than lower the risk of their portfolios – a
strategy that worked poorly during the global shock of the Great Depression.
Whereas the evidence using aggregated data suggests that branch banking
stabilizes banking systems (which, although not a direct test, could be viewed as
consistent with the diversification hypothesis), the studies using individual data point in
the other direction (and are inconsistent with the diversification view). The next section
therefore proposes an alternative hypothesis in an attempt to reconcile these conflicting
results.
Section 3. Branch Banking and Competition
During the 1920s, the number of branches operated by banks more than doubled,
increasing from 508 in 1920 to 1,131 in 1930 (Federal Reserve 1931, vol. 2).8 We
hypothesize that this expansion of branching networks increased the level of competition
in states that allowed branching to occur. As a result, banks that were only marginally
profitable prior to the increase in competition would not be profitable with the increase in
8
This represents an increase in the number of branches operated outside the city of the home office. Banks
also operated branches within the city of the home office. The number of these branches tripled between
1920 and 1930, rising to 2,391 at the end of the period. It is not clear that branches in the city of the home
office provided the same opportunities for diversification as branches outside the home office. Thus, in the
discussion of the hypothesis, the term branching will refer to the branches located outside the city of the
home office.
6
competition. In turn, these banks would likely merge with existing banks or voluntarily
liquidate.9 Also, because of the increased level of competition that branching induces, it
is likely that fewer banks would be able to find a profitable niche and enter the market.
With the exit of the weakest banks and less entry of new banks, the economic viability of
average bank would increase, and the rate of failure for banks within that state would
decline.
Why should these competitive forces apply to changes in branching laws and not
simply to new banks that emerged in the 1920s? First, banks with branches were more
cost-effective than new unit banks because jobs at different branches could be
consolidated and performed at the head office, reducing operating costs (Federal Reserve
1931, vol. 2, p. 224). Second, start up costs were lower for new branches than full service
banks, reducing barriers to entry. In some states, this included lower charter capital
requirements for branches than full-service banks (Southworth, 1928). Third, new
branches that were set up in previously restricted markets were likely more adept at
realizing higher rates of return than a comparable new unit bank since branches could
transfer deposits out of the local market to regions where capital was in higher demand.
The ability to obtain a cost advantage through branching and realize higher rates of return
thus made entry into existing local markets easier for branch banks than new unit banks.
The competitive forces that branch banking unleashed were not only limited to
large cities. Indeed, branch banking may have been instrumental in bringing banking and
banking competition to small towns. In 1931 (the only year for which we have been able
to locate the distribution of branches by town size), nearly half of all branches outside the
9
Alternatively the management could keep the bank open until it failed. This seems unlikely, however, as it
would involve a loss on the part of investors; selling or liquidating the bank prior to failure would not.
7
home-office city were located in towns of less than 2,500 people (Table 1). Figure 3
shows the locations of branches outside the home office. Moreover, the Federal Reserve
suggests that important rural banking networks (measured in terms of the size of banks
and share of the state’s total assets) existed in states such as North and South Carolina,
and these systems had been created, at least in part, through consolidation (Federal
Reserve, 1931, pp.214).
Laws that permitted statewide branching applied only to state-chartered banks
whereas the data we use to test the competition hypothesis are for national banks. The
logic of our hypothesis, however, still applies to national banks since they would be
subject to increased competition from the state banks that were allowed to establish
branches.10 Indeed, since the vast majority of national banks did not have branches, a
sample consisting of national banks may in fact be better for examining whether branch
banking affects competition because any observed or unobserved channels through which
branching might affect both competition and diversification are likely to be less of a
problem in a sample with fewer branched banks.11
It has been shown by Calomiris and Mason (2000), among others, that new banks
and less profitable banks were most likely to fail during the Great Depression. A key
component of our hypothesis is that fewer of these banks would exist in states that
allowed branch banking because the competitive pressures associated with the rise of
branch banking networks, prior to the Depression, would have forced weak banks to exit
10
Additionally, state banks that converted into national banks were permitted to keep branches they had
established while they were state banks, enabling branch banks to become national banks through a legal
technicality.
11
Excluding California, the country's nearly 8,000 nationally chartered banks operated 18 branches outside
the home-office city in 1925. By 1930 the number of branches had risen slightly to 27 (Federal Reserve
Board of Governors, 1943).
8
from the banking system and reduced the number of entries. In states without the
competitive pressures of branch banking, more weak banks would still exist at the start of
the Depression and these states would therefore have more banks that would be likely to
fail during the subsequent downturn. This interpretation is consistent with the findings in
Mitchener (2000a, 2000b) and Wheelock (1995): states allowing branch banking would
have had lower failure rates than those prohibiting it. And it would also be true even
though it was not necessarily the case that branch banks were the survivors.
The hypothesis that a banking system with branching is more diversified and
therefore more stable than a banking system with only unit banks presupposes that the
two systems are in equilibrium. When examining the 1920s and 1930s, however, it is
more consistent to consider an argument about what happens as the system is
transitioning from a unit banking equilibrium to a branch banking equilibrium. Our
hypothesis emphasizes how the expansion of branch banking within a state can increase
stability by pruning out inefficient banks through mergers and voluntary liquidations, and
is consistent with recent research examining the effects of the introduction of interstate
branching (Berger, Demsetz, and Strahan, 1999).
Our hypothesis stands in contrast to one of the longstanding populist arguments
lodged against branch banking. Opponents of branch banking have often complained that
it was a form of cartelization that would result in consolidation of the industry and
reduced competition, and that its growth would reduce the viability of businesses in small
communities by siphoning funds to urbanized areas. Such sentiments were widely
expressed in the first quarter of the 20th century – when branching was spreading
9
rapidly.12 While the spread of branch banking may lead to consolidation, the effects on
competition are not as clear as opponents of branching suggest. In fact, the economic
theory or private-interest view of regulation argues that branching restrictions are used to
protect inefficient, local monopolies and restrict competition.13 The result of these
intrastate regulations was less than full-scale competition in local deposit and loan
markets. As Chapman and Westerfield (1942, p.233) described the situation in the 1920s
and 1930s:
“Country bankers foresee danger to themselves in the possibility of
inroads into their areas of operation, should the larger institutions of the
cities be permitted to establish branches and compete with them in their
area on equal terms. They know that such a policy would result in a
reduction of interest rates in their towns and that their chances for the
profitable use of their funds might be somewhat diminished unless they
were prepared to go as far as their new rivals in serving customers
cheaply. The alleged apprehension of unit bankers as to the monopolistic
character of branch banking is, to say the least, selfish. What really
motivates them is their desire to preserve their local monopolies and
escape the competition of the more effective branch banks.”
Our hypothesis is sympathetic to this view of regulatory impediments. It is also
analogous to one that has been made in the context of global financial services industry:
just as foreign banks have brought necessary competition to inefficient domestic markets
(Folkerts-Landau and Lindgren, 1998), branch banking can introduce greater competition
to local markets, improving the cost and delivery of services to customers and the safety
of the banking system by forcing inefficient banks to merge or go out of business.
12
See, for example, the discussion in Chapman and Westerfield (1942, p.10) and Willit (1930) for
numerous examples of this view.
13
See Kroszner and Strahan (1998, 2000), Mitchener (2000c), and Chapman and Westerfield (1942).
10
Section 4. Testing the Competition/Consolidation Hypothesis
This section tests several predictions of our hypothesis using data on national
banks from the 1920s and the first two years of the Depression. First, we test whether the
number of mergers and voluntary liquidations was higher (and the number of entrants
lower) in states that permitted branch banking. We then test whether other factors related
to competition and consolidation including the number of banks per capita and the
profitability of different banks are related to branching regulation. Finally, we test
whether the “exit/no entrance” rate better explains failure rates than simply whether
branch banking was allowed. To implement this last test, we start by reproducing the
standard aggregate data result that statewide branching is negatively associated with bank
failures. We then construct an index that measures the extent of consolidation in the
state’s bank system. The index is then used in the failure rate regression and we test
whether it is significant and whether the indicator for being a state that allows branch
banking becomes insignificant.
A. Mergers, Voluntary Liquidations, and Entry in Branch Banking States
Our first test examines industry consolidation – whether there were more mergers,
voluntary liquidations, and fewer entries for national banks in states that allowed branch
banking.14 Table 2 summarizes the data on bank entries and exits for national banks over
14
Wheelock (1992, p.815) suggests that this may have occurred in the 1920s, but does not formally test this
notion: “Like most Midwestern states, Kansas was a unit banking state during the 1920s, with over 1,000
small banks in operation…Had branching restrictions been removed those counties might have experienced
greater consolidation, through either mergers or failures.”
11
the sample period of 1921-30, and groups the data by branching law. (Appendix Table 1
describes the sources for our data.) As the last column of the table shows, during the
1920s, states permitting statewide branching averaged more mergers and voluntary
liquidations and had fewer new entries into the banking system than states limiting
branching (to the city or county level), prohibiting branching, or having no law with
respect to branching. The results are most striking for mergers and voluntary liquidations.
To explore these results further, we regress the number of exits and entries for
national banks on dummies for the prevailing law regarding branching and year dummies
over our full sample as well as subperiods. Based on the states’ laws during the previous
year, the states are classified into one of four categories: those prohibiting branching,
those allowing limited branching (generally branches within the city or county of the
home office), those allowing statewide branching, and those having no laws with respect
to branching. In states where the de jure situation differs from the de facto situation
(such as West Virginia where the laws allowed branching but the state banking
commissioner refused all applications to establish branches), the de facto situation is used
instead of the de jure. Detailed information on state branching laws (including whether
the de jure situation was used) is shown in Appendix Table 2.
Since exits and entries take on only discrete integer values, and are bounded
below by zero, we analyze the number of exits and entries employing count data analysis
rather than using other estimation procedures such as ordinary least squares.15 Moreover,
15
Count data analysis assumes that the dependent variable was generated through a Poisson process.
Determining the effect of the independent variables involves maximizing the log-likelihood function
∑ {y X β − exp(X β ) − ln y !}where y is the dependent variable vector, X is the matrix of
n
i =1
i
'
i
'
i
i
independent variables, β is the coefficient vector, and n is the number of observations. When analyzing
counts, it is important to control for the size of the population from which the counts are produced. In the
12
since entries and exits are truncated at zero rather than censored at zero, using Tobit
would produce biased estimates.16 For exits, we find that a Poisson distribution provides
a good fit; however, for entries, the variance greatly exceeds the mean, so we correct for
this overdispersion in the data using a negative binomial distribution. The coefficients
can be interpreted as the percentage change in the dependent variable (number of entries
or exits divided by the log of the number of banks in the state) due to a one-unit change
in the independent variable.
The estimated coefficients support the hypothesis that branch banking leads to
increased mergers and consolidation in a state’s banking system. During the 1920s, states
that allowed branch banking had more voluntary liquidations than other states (Table 3).
The evidence is strongest during the first half of the decade and is significant over the full
sample period. The point estimate for the full sample period suggests that states
permitting branch banking had 33 percent more voluntary liquidations per log bank than
states where branching was prohibited.
Statewide branching appears also to have impacted mergers significantly during
the 1920s. The count data analysis displayed in Table 4 shows that states permitting
unrestricted branching had more mergers during the 1920s. The coefficient on the
statewide branching variable over the entire sample period suggests that states permitting
statewide branching had 31 percent more mergers per log bank each year than those
prohibiting it. Finally, Table 5 shows that states permitting branching networks had lower
Poisson distribution the rate of “arrival” is constant so that the number of arrivals depends on the
population size. Increasing the population by a certain percent should increase the number of arrivals by the
same percent. In the estimation procedure above, this is equivalent to imposing the constraint that the effect
of the log of the population is equal to 1. For a detailed treatment of count data analysis see Cameron and
Trivedi (1998).
16
See Plassman and Tideman (2001).
13
rates of entry during the 1920s, around half the number of entries per log bank than states
forbidding branching. However, in contrast to the findings for voluntary liquidations, the
effect of statewide branching on entry is stronger during the second half of the 1920s than
the first half.17 (The number of entries per log bank in statewide branching states during
the second half of the decade was only about one-third that of other states.) This may
reflect the fact that the differences in mean rates of entry based on branching laws are less
pronounced in the 1920s compared to either voluntary liquidations or mergers (as shown
in Table 2). Or, it may reflect a response by entrepreneurs to seek out new markets
(where previously high profits had been generated by geographical monopolies) before
significant consolidation occurred (driven by mergers and voluntary liquidations) as a
result of branching.
Another more general way of testing the effects that branch banking had on
consolidation is to examine whether states with branch banking laws had fewer banks per
capita and whether these states had larger declines in the number of banks over the
decade. The first two columns of Table 6 present results from regressing banks per capita
on indicators of state branching laws and year dummies. The OLS estimates show that
states permitting statewide branching had fewer banks per capita during the 1920s (Table
6). The next two columns of Table 6 show a regression of the percentage change in the
number of banks over the decade on branching laws. Again, we find that states permitting
statewide branching had more consolidation. These results are consistent with previous
research, which has shown that the percentage change in banks per capita in the 1920s
17
The shift in the effects of branch banking on exits and entrances coincides with the shift in bank failures
patterns observed by Allston, Grove, and Wheelock (1994).
14
was greatest in states where branching expanded the most (Wheelock, 1993).18 Because
we additionally find that branch banking tends to increase mergers and voluntary
liquidations, we are fairly confident that the reduction in the number of banks per capita
in a state was caused by branch banking.
B. Consolidation and Increased Competition versus Monopoly Power
Although the results presented in the previous section support the view that
branch banking increases consolidation in the banking industry, it is unclear whether this
consolidation led to a more competitive banking system or a monopolistic banking
system. For example, when banks are forbidden from establishing branches, it artificially
segments the market and enables unit banks to develop monopolies within particular
localities. The immediate impact of branching is that it introduces competition into these
geographically segmented markets. If banks that were previously insulated from the cool
winds of competition are acquired or forced out of the market as a result of this increased
competition, it is possible that the surviving banks in turn acquire monopoly power. This
was one of the concerns of those who lobbied against more permissive branching laws in
the 1920s: the end result of liberalization would simply be further concentration in the
banking industry. To investigate further how statewide branching affects the competitive
environment of a state’s banking system, we examine the relationship between branching
and two additional measures: banking sector concentration and bank profits.
18
Our results are also consistent with Berger, Kashyap, Scalise (1995), who show that the introduction of
nationwide banking beginning in the 1980s accelerated the reduction in the number and market share of
small bank organizations for the period of recent regulatory liberalization.
15
We first test whether states allowing branch banking had more concentrated
banking sectors by regressing bank concentration on indicators of the state banking laws
and year dummies for two periods: 1921-1930 and 1926-1930.19 The first measure of
concentration is essentially a Herfindahl index, constructed by computing the share of
national banks in a particular size category (based on loans and investments) and
weighting the size categories by loan and investment volume. We use the mean of the
size category of national banks as the value of the national bank’s size rather than the
national bank’s actual size, which is unknown. Data on bank size (described in Appendix
Table 1) are only available for 1926-30, so the results for this measure of banking are
presented for this period. The second measure is a four-firm concentration ratio based on
the largest four national banks in each state for each year of our sample. We calculate this
measure for 1921-1930. Using either measure, our OLS estimates show that states
permitting branch banking tended to have more concentrated banking sectors (Table 7).
Next, we examine how branch banking affects bank profits. Since national banks
were all regulated at the national level by the Office of the Comptroller of Currency and
were largely restricted in their ability to have branches, differences in the level of
competition that they faced would be a significant factor in creating differences in profit
levels between states. Finding lower profits in states that allowed branch banking would
indicate that the national banks faced heavier competition from state banks in these
locations as a result of branching.20
19
The sample periods are constrained by the availability of data for constructing the two measures of
banking industry consolidation.
20
Lower profitability may also result from higher merger costs; however, due to branching restrictions on
nationally-chartered banks, this interpretation is more likely to apply to state banks than the national banks
in our sample. Alternatively, national banks may have higher profits if branching leads to so much industry
consolidation that the surviving banks are able to generate oligopoly profits. Higher profits may also result
if competition has forced a sufficient number of inefficient national banks from the banking system.
16
To test how branching laws affect profitability, we use profit data on national
banks that are available from 1926-30 and that are decomposed by both size category
(based on loans and investments) and by state. The data are grouped into seven profit
intervals, using the percentage return to capital as the measure of profitability.21 Because
the categories are ordered but the distance between them is not constant, an ordered logit
is estimated using weights equal to the number of banks in each size category earning a
given level of profits. The independent variables are bank size and dummies indicating
the states’ laws regarding branch banking.
Table 8 shows that states allowing statewide branch banking tended to have lower
profits than states prohibiting it. The signs on the statewide branching coefficients shown
in tables 7 and 8 suggest that branching appears to have increased concentration, but not
to the point where branch banking networks acquired monopoly power. Our findings are
also consistent with Berger, Kashyap, and Scalise (1995), who view the removal of
geographic restrictions as likely reducing the exercise of market power (by unleashing
more actual or potential customers into local markets) and improving allocative
efficiency (by enabling resources to flow more easily toward activities that yielded higher
returns and more efficient producers).
C. Branch Banking and Financial Stability
We now test whether the consolidation caused by statewide branching is
responsible for the lower failures found in states allowing branch banking. We construct
21
These categories are: deficit of 6% or more, deficit of 5.9% to 0 %, profit of less than 3%, profit of 3% to
5.9%, profit of 6% to 8.9%, profit of 9% to 11.9%, and profit of 12% and over.
17
an annual index of consolidation to measure the degree to which inefficient banks have
exited a state’s banking system and then observe how the index affects a regression of
bank failures on branching.
Our consolidation index is a weighted sum of the fraction of banks in state i that
experience voluntary liquidations, L, and mergers, M, minus the share of banks entering
the system, E. The index takes the form:
t −1
(1) C it = ∑ [ wLt Lit + wMt M it − wEt Eit ] .
t −4
Failures are not included as we assume that they are not directly related to
consolidation.22 Because the index for year t is meant to capture consolidation that
occurred in previous years, the index is comprised of entries and exits during the previous
four years.23 Different weights are assigned to the number of entries per year and to the
number of exits per year. These weights reflect the view (and our preliminary findings in
Section 4A) that branch banking may have initially induced more mergers and voluntary
liquidations as shake out in the industry occurred early in the decade, and then
discouraged entry as the decade preceded. Thus exits earlier in the decade receive more
weight than current exits, and more recent entrants receive more weight than those that
occurred four years prior. These results, however, are robust to a variety of weights.24
To test whether consolidation in a state’s banking sector, induced by branch
banking, improves the stability of banking systems, we analyze the relationship between
22
23
For this reason, we also do not simply use the change in the number of banks.
Similar results are obtained as long as three or more prior years are included in the index.
18
branch banking and failure rates with and without the inclusion of this index. We
examine bank failures for 1926-1930, which are indexed by state and size category over
this sample period. Our data thus permit us to control for bank size – a factor that has
been associated with the probability of failures in previous studies (Calomiris and Mason
2000, Carlson 2001, White 1984).25 Figure 4 shows the distribution of banks and failures
by bank size category. We estimate the number of failures using count data analysis,
adjusting for the number of banks in the state. We drop bank size/state categories that
contain only 1 or 2 banks, as including these categories increase the noise of the
estimates.
Estimates using bank size/state categories with 1 or 2 banks are still
significant at conventional levels, though to a lesser degree.
We include the branching regime, bank size dummies, year dummies, and
measures for agricultural and business sector distress to control for differences in real
shocks across states. In particular, to capture differences in business conditions across
states, a measure of lagged business failures is computed. We include two variables to
capture state-level differences in agricultural conditions as much of the banking distress
in the 1920s was associated with predominantly agricultural states: (1) an index of lagged
farm real estate farm prices and farm foreclosure (bankruptcies) rates. Following
Wicker’s (1996) finding that the bank panic of 1930 was a regional event caused by the
failures of Caldwell and Company, we also include an indicator variable for whether the
state had banks failures associated with the failure of Caldwell and Company in 1930.
24
The weights used are wL,t-1 = wM,t-1 = 0.25, wL,t-2 = wM,t-2 = 0.5, wL,t-3 = wM,t-3 = 1, wL,t-4 = wM,t-4 = 1,
wE,t-1 = 1, wE,t-2 = 4, wE,t-3 = 4, wE,t-4 = 1.
25
Larger banks may be more stable because of their greater ability to spread risks, coordinate emergency
assistance with other banks, and reduced propensity to spread contagion when the banking sector is
subjected to external shocks.
19
The econometric results support the hypothesis that branch banking improves
financial stability by weeding out inefficient banks through increased competition and
consolidation. When the index of bank consolidation is not included (Table 9), states
permitting branch banking had lower failure rates (about 50 percent fewer failures per log
bank). This negative relationship is similar to the results shown in previous studies using
aggregate data on bank failures, and exists even when differences in the size of banks are
included in the regression. However, when the index of bank consolidation is included as
an additional covariate, then branch banking no longer has any predictive power
regarding failure rates. This finding has two important implications. First, diversification
cannot explain the improved stability of state banking systems; otherwise, the coefficient
should still be statistically significant. Second, the fact that the consolidation index is
negatively associated with failures (an increase in the index of one standard deviation
reduces the number of failures per log bank by 22 percent) and is statistically significant
suggests that consolidation in the banking industry (which occurred to a greater extent in
states permitting branch banking) stabilized state banking systems in the 1920s and
1930s. This relationship persists and is robust to the inclusion of differences in real
factors, including agricultural distress, which has previously been identified as the main
determinant of bank failures in the 1920s.26 Moreover, our results confirm what some
economic historians have suggested, but not shown. As White (1985) contends, “The
number of small banks in rural areas needed to be reduced, and the mergers assisted the
weaker institutions with less pain than the massive failures that followed. Unfortunately,
this development was stifled by regulations in most states that forbade branch banking.”
26
Alston, Grove, and Wheelock (1994) identified agricultural distress as the most important factor in
explaining differences in bank failures across states during the 1920s.
20
D. Testing the Role of Mergers
Another possibility is that branching decreases failures by facilitating an
alternative to failure, namely merging with another bank. Branching facilitates mergers
by expanding the pool of possible merger partners from banks within a city to all statechartered banks within the state. We test this hypothesis by constructing an index of the
ease with which banks might merge. This index is the number of mergers divided by the
total number of bank exits (mergers, failures, and voluntary liquidations). When the
lagged ease-of-merger index is included in the bank failure regression without the
consolidation index (Table 10), it is not significant and the statewide branch-banking
indicator remains significant. When both the ease-of-merger index and the consolidation
index are included, the consolidation index is statistically significant and negative in sign
while the ease-of- merger index is still not significant. Moreover, the branch-banking
indicator becomes insignificant as it did in our previous regressions when the
consolidation index was included. This suggests that, although the increased pool of
prospective merger partners may have played a role in reducing bank failures, the
consolidation of the banking sector produced by branching was more important.
Section 5. Conclusion
This paper revises our understanding of how branching affects financial stability.
We argue that states that permitted statewide branching experienced lower failure rates
21
during the 1920s and the Depression, not because branch banks were more diversified,
but because branching laws transformed the competitive environment. Branching laws
increased the level of competition by breaking up local geographic monopolies and
encouraging consolidation by forcing inefficient banks to merge or exit the system. Using
data on national banks from the 1920s, we find that branch-banking states had more
voluntary liquidations and mergers, which in turn led to lower failure rates during this
period. Branch banking was therefore positively associated with improved systemic
stability, but not for the reasons that have been traditionally emphasized.
The well-documented response of unit bankers in the 1920s to the growth of
branch banking was to lobby state and federal governments to legally limit it from
spreading. This behavior appears consistent with the evidence presented here and with
the economic theory of regulation. Branching was increasing competition in states where
it was permitted and eroding the monopoly profits of unit bankers that had previously
benefited from an entry barrier that had been removed.
22
References:
Allen, Franklin and Douglas Gale. (2000). Comparing Financial Systems. Cambridge,
MA: MIT Press.
Alston, Lee, Wayne Grove, and David Wheelock. (1994). “Why Do Banks Fail?
Evidence from the 1920s.” Explorations in Economic History, Vol. 31, pp. 409-31.
Berger, Allen, Rebecca Demsetz, and Phillip Strahan. (1999). “The consolidation of the
Financial Services Industry: Causes, Consequences, and Implications for the Future.”
Journal of Banking and Finance, Vol. 23.
Berger, Allen, Anil Kashyap, and Joseph Scalise. (1995). “The Transformation of the
U.S. Banking Industry: What a Long, Strange Trip It’s Been,” Brookings Papers on
Economic Activity, Issue 2, pp. 55 – 218.
Calomiris, Charles. (2000). U.S. Bank Deregulation in Historical Perspective.
Cambridge: The Cambridge University Press.
Calomiris, Charles and Joseph Mason. (2000). “Causes of U.S Bank Distress During the
Depression,” NBER Working Paper 7919.
Cameron, A. Colin and Pravin Trivedi. (1998). Regression Analysis of Count Data.
Cambridge: The Cambridge University Press.
Carlson, Mark. (2001). “Are Branch Banks Better Survivors? Evidence from the
Depression Era,” Finance and Economics Discussion Series 2001-51.
Chapman, John and Ray Westerfield. (1942). Branch Banking: Its Historical and
Theoretical Position in America and Abroad, New York: Harper and Brothers Publishers.
Cherin, Antony and Ronald Melicher. (1988). “Impact of Branch Banking on Bank Firm
Risk via Geographic Market Diversification,” Quarterly Journal of Business and
Economics Vol. 27(2), pp. 73-95.
Comptroller of the Currency. (various years). Annual Report. Washington D.C.: United
States Government Printing Office.
Comptroller of the Currency. (various years). Statements of National Banks. Washington
D.C.: United States Government Printing Office.
Drummond, Ian. (1991). “Why Canadian Banks Did Not Collapse in the 1930s.” In The
Role of Banks in the Interwar Economy, edited by James, et al. Cambridge: Cambridge
University Press.
23
Federal Reserve Board of Governors (1931). Report of the Branch, Chain, and Group
Banking Committee Volumes 2 and 9. Washington D.C.
Federal Reserve Board of Governors (1943). Banking and Monetary Statistics.
Washington D.C.: The National Capital Press.
Friedman, Milton and Anna Schwartz. (1963). A Monetary History of the United States,
1867-1960. Princeton: Princeton University Press.
Gart, Alan. (1994). Regulation, Deregulation, Reregulation. New York: John Wiley and
Sons, Inc.
Grossman, Richard. (1994). “The Shoe That Didn’t Drop: Explaining Banking Stability
in the Great Depression,” Journal of Economic History, Vol. 54, pp. 654-682.
Hubbard, R. Glenn. (1994). Money, the Financial System, and the Economy. Reading,
MA: Addison-Wesley Publishing Company.
Folkerts-Landau, David and Carl-Johan Lindgren. (1998). Toward a Framework for
Financial Stability. Washington: International Monetary Fund.
Kroszner, Randall and Phillip Strahan. (1998). “What Drives Deregulation? Economics
and Politics of the Relaxation of Bank Branching Restrictions.” Quarterly Journal of
Economics, Vol. 114(3), pp. 1437-67.
Kroszner, Randall and Phillip Strahan. (2000). “Obstacles to Optimal Policy: The
interplay of Politics and Economics in Shaping Bank Supervision and Regulation
Reforms.” NBER Working Paper 7582.
Kryzanowski, Lawrence and Gordon Roberts. (1993). “Canadian Banking Solvency,”
Journal of Money Credit and Banking, Vol. 25, pp. 361-376.
Lauch, Louis and Neil Murphy. (1970). “A Test of the Impact of Branching on Deposit
Variability.” Journal of Financial and Quantitative Analysis (September), Vol. 5, pp. 323-27.
Mitchener, Kris. (2000a). “Do Supervision and Regulation Influence Financial Stability?
Evidence from the Great Depression,” unpublished manuscript, University of California,
Berkeley.
Mitchener, Kris. (2000b). “Bank Supervision, Regulation, and Financial Stability During
the Great Depression,” unpublished manuscript, University of California, Berkeley.
Mitchener, Kris. (2000c). “Regulatory Chaos and the Great Depression: The Evolution of
State Bank Regulation,” unpublished manuscript, University of California, Berkeley.
24
Plassmann, Florenz and T. Nicolaus Tideman. (2001). “Does the Right to Carry
Concealed Handguns Deter Countable Crimes? Only a Count Analysis Can Say,” Journal
of Law and Economics, Vol. 34, pp.771-98.
Polk’s Bankers Encyclopedia Co. (various years). Polk’s Bankers Encyclopedia. Detroit,
MI: Polk’s Bankers Encyclopedia Co.
Preston, Howard. (1928). “Recent Developments in Branch Banking,” American
Economic Review, Vol. 14, pp. 443-62.
Rose, Peter and William Scott. (1978). “Risk in Commercial Banking: Evidence from
Postwar Failures.” Southern Economic Journal (July), pp. 90-106.
Southworth, Shirley. (1928). Branch Banking in the United States, New York, New
York: McGraw-Hill Book Company.
Study Commission for Indiana Financial Institutions. (1932). Report. Indianapolis, Ind.:
Seventy-Seventh General Assembly.
U.S. Department of Agriculture. (1936). Circular No. 414 Farm Bankruptcies, 18981935. Washington: U.S. Department of Agriculture.
U.S. Department of Agriculture. (1939). Circular No. 548 The Farm Real Estate
Situation. Washington: U.S. Department of Agriculture.
U.S. Department of Commerce. (various years). Statistical Abstract of the United States.
Washington, DC: GPO.
Wacht, Richard. (1968). “Branch Banking and Risk.” Journal of Financial and
Quantitative Anaylsis, Vol. 3(March), pp. 97-107.
Wheelock, David. (1993). “Government Policy and Banking Market Structure in the
1920s.” Journal of Economic History, Vol. 53(4), pp. 857-879.
Wheelock, David. (1995). “Regulation, Market Structure, and the Bank Failures
of the Great Depression,” Federal Reserve Bank of St. Louis Review, Vol. 77, pp.
27-38.
White, Eugene. (1983). The Regulation and Reform of the American Banking System,
1900-1929. Princeton, N.J.: Princeton University Press.
White, Eugene. (1984). “A Reinterpretation of the Banking Crises of 1930,”
Journal of Economic History, Vol. 44, pp. 119-138.
White, Eugene. (1985). “The Merger Movement in Banking, 1919-33,” Journal of
Economic History, Vol. 45(2), pp.285-91.
25
Wicker, Elmus. (1996). The Banking Panics of the Great Depression. Cambridge:
Cambridge University Press.
26
Figure 1
Branches of U.S. Banks Located Outside the Head-Office
City
Number of Branches
1200
1000
800
600
400
200
0
1900
1905
1910
1915
Year
Source: Federal Reserve Board (1931), vol. 2
27
1920
1925
1930
Figure 2
Bank Failures in the United States
1600
Number of Failures
1400
All Banks
National Banks
1200
1000
800
600
400
200
Year
Source: Federal Reserve Board (1943)
28
30
19
29
19
28
19
27
19
26
19
25
19
24
19
23
19
22
19
19
21
0
Figure 3 – Branches of National and State Banks outside the City of the Head Office (December 31, 1931)
Source: Federal Reserve Board (1931, Vol. 2, p.17)
Note: In California there are numerous branches in the metropolitan areas centering around San Francisco
and Los Angeles, but technically outside their city limits. On the map the dots extend far beyond the
territory in which the branches are actually located around these cities.
29
Figure 4 – Annual Average Number of National Banks and Bank Failures by Size Category
1800
Banks
1600
Failures
1400
1200
1000
800
600
400
200
0
1
2
3
4
5
6
7
8
9
10
Size Categories
Sources and Notes: Comptroller of the Currency (various years) and Federal Reserve Board (1931). Annual
averages are constructed over the period 1926-1930. Size categories are: category 1 - under $150,000;
category 2 - $150,000-$250,000; category 3 - $250,000-$500,000; category 4 - $500,000-$750,000;
category 5 - $750,000-$1,000,000; category 6 - $1,000,000-$2,000,000; category 7 - $2,000,000$5,000,000; category 8 - ($5,000,000-$10,000,000; category 9 - $10,000,000-$50,000,000; category 10 $50,000,000 and over.
30
Table 1 – The Distribution of Bank Branches by Town Population (1931)
Town Population
Under 500
500-1,000
1,000-2,500
2,500-5,000
5,000-10,000
10,000-25,000
25,000-50,000
50,000-100,000
100,000+
Total
Branches Outside Branches Inside Home
All Branches
Home Office City
Office City
Percent of
Percent of
Percent of
Number
Number
Number
Total
Total
Total
189
16.32
2
0.09
191
5.73
173
14.94
0
0.00
173
5.19
207
17.88
7
0.32
214
6.42
134
11.57
7
0.32
141
4.23
107
9.24
9
0.41
116
3.48
91
7.86
27
1.24
118
3.54
46
3.97
63
2.90
109
3.27
60
5.18
132
6.07
192
5.76
151
13.04
1929
88.65
2080
62.39
1158
100
2176
Source: Federal Reserve Board (1931), vol. 2.
31
100
3334
100
Table 2 – Summary Statistics for National Banks
Number of states
Average number of
banks
No Law Average voluntary
liquidation rate
Average merger rate
Average entry rate
Number of states
Average number of
banks
Average voluntary
Unit
liquidation rate
Average merger rate
Average entry rate
Number of states
Average number of
banks
Limited
Average voluntary
Branching
liquidation rate
Average merger rate
Average entry rate
Number of states
Average number of
banks
Statewide
Average voluntary
Branching
liquidation rate
Average merger rate
Average entry rate
1921
11
1922
11
1923
11
1924
9
1925
9
1926
9
1927
9
1928
7
1929
6
1930 Average
5
8.7
175.9 188.4 181.7 159.3 153.3 143.0 138.4 142.6 154.3 121.6
0.8
0.7
1.3
19
4.5
3.5
3.1
23
2.1
2.0
1.7
20.3
167.0 168.5 168.3 168.8 171.4 161.3 151.7 143.5 139.2 141.9
157.2
0.8
0.4
2.0
7
0.9
1.2
2.8
19
1.7
1.2
2.1
7
1.7
1.9
2.2
19
2.6
2.2
1.4
19
1.8
2.0
1.5
7
1.4
1.6
1.6
9
2.1
2.0
1.7
19
2.6
2.3
2.6
9
1.8
1.8
1.2
19
1.9
2.5
3.1
9
2.2
2.4
1.1
20
2.0
2.2
1.2
9
1.5
1.8
0.7
23
2.5
2.3
1.5
9
5.2
5.1
2.0
23
159.6
3.1
3.1
2.4
10
4.8
4.1
1.9
11
2.3
2.2
1.9
8.7
281.9 282.3 284.6 268.4 272.2 267.8 269.2 269.3 243.2 225.5
264.2
2.3
1.2
2.3
11
1.1
1.0
1.2
11
0.9
0.7
2.6
11
1.9
1.7
3.7
11
0.5
0.6
2.4
11
1.3
1.2
2.7
11
1.2
1.3
1.4
10
1.8
1.9
2.2
9
1.9
1.9
2.0
9
2.2
2.0
1.3
9
1.6
1.4
2.2
10.3
97.5
96.2
95.0
92.4
91.6
85.1
86.0
84.1
80.1
74.4
88.8
1.0
0.6
1.7
2.2
1.6
3.6
1.9
2.2
1.6
3.1
3.4
1.4
2.5
1.6
2.6
5.1
4.1
1.1
2.7
2.9
0.9
3.8
4.4
0.7
6.3
7.2
2.0
8.2
6.0
0.2
3.6
3.3
1.6
Sources and Notes: Rates are computed using data from the Annual Report of the Comptroller of the
Currency (1921-1930). Rates are expressed as percentages of existing banks. Within category rates are
unweighted averages of state rates. The last column shows the average of the yearly observations over the
entire sample period, 1921-1930.
32
Table 3 – Effect of Branching on Voluntary Liquidations.
Dependent variable: The count of voluntary liquidations in a state per year divided by the log of the
number of banks in the state.
Full Sample
Early 1920s
Late 1920s
(1921-1930)
(1921-1926)
(1925-1930)
Intercept
Population
No Law
(
(
(
Statewide
Branching
(
Limited
Branching
(
1922
1923
1924
1925
1926
1927
1928
1929
1930
(
(
(
(
(
(
(
(
(
-1.84 ***
0.27 )
0.24 ***
0.03 )
0.69 ***
0.16 )
0.33 **
0.15 )
-0.94 ***
0.21 )
0.16
0.33 )
0.30
0.32 )
0.62 **
0.31 )
0.41
0.32 )
0.63 **
0.31 )
0.67 **
0.30 )
0.65 **
0.31 )
1.01 ***
0.29 )
1.22 ***
0.28 )
(
(
(
(
(
(
(
(
(
(
-1.89 ***
0.30 )
0.25 ***
0.04 )
0.66 ***
0.23 )
0.46 **
0.22 )
-0.83 ***
0.32 )
0.16
0.33 )
0.30
0.32 )
0.61 **
0.31 )
0.39
0.32 )
0.61 **
0.31 )
(
(
(
(
(
(
(
(
(
(
-1.44 ***
0.24 )
0.25 ***
0.03 )
0.81 ***
0.20 )
0.25
0.18 )
-1.08 ***
0.25 )
0.23
0.28 )
0.27
0.27 )
0.26
0.28 )
0.62 **
0.26 )
0.83 ***
0.25 )
Observations
480
288
288
Log-Likelihood -374.8
-218.2
-231.4
Mean dependent variable: 0.62
Mean voluntary liquidations: 3.2
Mean log(banks in operation): 4.6
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Standard errors are in parentheses. Data on voluntary liquidations are from the Annual Report
of the Comptroller of the Currency (1921-1930). Data on branching laws are from the Federal Reserve
Board (1931), and population is from the 1930 Census. The omitted branching group is states that
prohibited branching. The first year in each sample period is the omitted year from the regression.
Observations in the ratio regression are weighted by state population.
33
Table 4 – Effects of Branching on Mergers
Dependent variable: The count of mergers in a state per year divided by the log of the number of banks in
the state.
Full Sample
Early 1920s
Late 1920s
(1921-1930)
(1921-1926)
(1925-1930)
Intercept
Population
No Law
Statewide
Branching
Limited
Branching
1922
1923
1924
1925
1926
1927
1928
1929
1930
Count
-2.36 ***
( 0.34 )
0.25 ***
( 0.03 )
0.72 ***
( 0.17 )
0.31 **
( 0.16 )
-0.98 ***
( 0.22 )
0.66 *
( 0.39 )
0.96 ***
( 0.37 )
1.08 ***
( 0.37 )
0.84 **
( 0.38 )
1.13 ***
( 0.37 )
1.24 ***
( 0.36 )
1.15 ***
( 0.37 )
1.59 ***
( 0.35 )
1.53 ***
( 0.35 )
Count
-2.40 ***
( 0.36 )
0.26 ***
( 0.04 )
0.71 ***
( 0.23 )
0.36
( 0.23 )
-0.96 ***
( 0.34 )
0.66 *
( 0.39 )
0.96 ***
( 0.37 )
1.08 ***
( 0.37 )
0.83 **
( 0.38 )
1.12 ***
( 0.37 )
Count
-1.52 ***
( 0.25 )
0.25 ***
( 0.03 )
0.78 ***
( 0.21 )
0.23
( 0.19 )
-1.05 ***
( 0.25 )
(
(
(
(
(
0.29
0.28 )
0.41
0.28 )
0.32
0.28 )
0.76 ***
0.26 )
0.70 ***
0.26 )
Observations
480
288
288
Log-Likelihood --367.6
-209.2
-232.1
Mean dependent variable: 0.59
Mean number of mergers: 3.1
Mean log(banks in operation): 4.6
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Standard errors are in parentheses. Data on mergers are from the Annual Report of the
Comptroller of the Currency (1921-1930). Data on branching laws are from the Federal Reserve Board
(1931) and population is from the 1930 Census. The omitted branching group is states that prohibited
branching. The first year in each sample period is the omitted year from the regression. Observations in the
ratio regression are weighted by state population.
34
Table 5 – Effect of Branching on Entry
Dependent variable: The count of new banks in a state per year divided by the log of the number of banks
in the state.
Full Sample
(1921-1930)
Intercept
Population
No Law
Statewide
Branching
Limited
Branching
1922
1923
1924
1925
1926
1927
1928
1929
1930
Dispersion
-1.10 ***
( 0.23 )
0.26 ***
( 0.03 )
0.27
( 0.18 )
-0.34 *
( 0.19 )
-0.46 **
( 0.20 )
0.31
( 0.26 )
0.07
( 0.27 )
-0.21
( 0.29 )
0.33
( 0.26 )
-0.09
( 0.29 )
-0.34
( 0.30 )
-0.47
( 0.31 )
-0.20
( 0.29 )
-0.44
( 0.31 )
0.23
( 0.07 )
Early 1920s
(1921-1926)
Late 1920s
(1925-1930)
-1.22 ***
( 0.26 )
0.27 ***
( 0.04 )
0.37 *
( 0.22 )
-0.18
( 0.24 )
-0.38
( 0.26 )
0.32
( 0.28 )
0.07
( 0.29 )
-0.22
( 0.31 )
0.33
( 0.28 )
-0.09
( 0.30 )
-0.63 ***
( 0.21 )
0.26 ***
( 0.03 )
-0.10
( 0.29 )
-0.61 **
( 0.27 )
-0.63 **
( 0.25 )
(
(
(
(
(
0.36
0.10
(
-0.42
0.26 )
-0.67 **
0.28 )
-0.82 ***
0.29 )
-0.56 **
0.27 )
-0.82 ***
0.29 )
0.18
0.09 )
Observations
480
288
288
Log-Likelihood -350.7
-210.5
-199.9
Mean dependent variable: 0.62
Mean number of entries: 3.4
Mean log(banks in operation): 4.6
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Standard errors are in parentheses. Data on new banks are from the Annual Report of the
Comptroller of the Currency (1921-1930). Data on branching laws are from the Federal Reserve Board
(1931), and population is from the 1930 Census. The omitted branching group is states that prohibited
branching. The first year in each sample period is the omitted year from the regression. Observations in the
ratio regression are weighted by state population.
35
Table 6 – Branching and the Dispersion of Banks
Banks Per Capita
(Sample Period: 1921-1930)
Percent Change in the
Number of Banks
(Sample Period: 1922-1930)
Coefficient
Estimate
0.93 ***
S.E.
( 0.06 )
Coefficient
Estimate
0.02
S.E.
( 0.01 )
0.73 ***
( 0.05 )
-0.01
( 0.01 )
Statewide Branching
-0.30 ***
( 0.04 )
-0.02 *
( 0.01 )
Limited Branching
-0.21 ***
( 0.04 )
Intercept
No Law
0.02 **
( 0.01 )
1922
0.00
( 0.07 )
1923
-0.04
( 0.07 )
-0.03 **
( 0.01 )
1924
-0.09
( 0.07 )
-0.04 ***
1925
-0.12 *
( 0.07 )
-0.03 *
( 0.01 )
( 0.01 )
1926
-0.17 **
( 0.07 )
-0.06
( 0.01 )
1927
-0.20 ***
( 0.07 )
-0.03 **
( 0.01 )
1928
-0.23 ***
( 0.07 )
-0.03 **
( 0.01 )
1929
-0.26 ***
( 0.07 )
-0.05 ***
( 0.01 )
1930
-0.29 ***
( 0.07 )
-0.05 ***
( 0.01 )
Observations
431
476
Adjusted R-square
0.05
0.35
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Estimated using ordinary least squares. Information on the number of bank and on branching
laws is from the Federal Reserve Board (1931), and population is from the 1930 Census. The omitted
branching group is branches prohibited. The first year in each time period is the omitted year.
36
Table 7 – Branching and Industry Concentration
Dependent variable: concentration indexes
Herfindahl Index
(Sample: 1926-1930)
Four-firm
Concentration Index
(Sample: 1921-1930)
Variable
Coefficient
Estimate
S.E.
Coefficient
Estimate
Intercept
1543.60 ***
( 33.0 )
0.39 ***
( 0.02 )
Population
-121.53 ***
( 5.8 )
-0.02 ***
( 0.00 )
No Law
120.44 ***
( 33.9 )
-0.13 ***
( 0.02 )
Statewide Branching
217.64 ***
( 35.5 )
0.05 ***
( 0.02 )
-550.45 ***
( 39.1 )
0.01
( 0.02 )
1922
0.04
( 0.03 )
1923
-0.01
( 0.03 )
1924
0.00
( 0.03 )
1925
0.02
( 0.03 )
Limited Branching
S.E.
1926
-45.19
( 39.4 )
0.03
( 0.03 )
1927
30.29
( 39.4 )
0.05 *
( 0.03 )
0.06 *
( 0.03 )
1928
1929
50.3919
( 39.3 )
0.05 *
( 0.03 )
1930
98.6049 **
( 39.3 )
0.09 ***
( 0.03 )
Observations
2399
479
Adjusted R-square
0.23
0.17
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Estimated using ordinary least squares. Information on the number of bank, size distribution
and on branching laws is from the Federal Reserve Board (1931). Four-firm concentration index derived
from Polk’s Bank Directory (various years) and the Comptroller of the Currency (various years).
Population is from the 1930 Census. The omitted branching group is branches prohibited. The first year in
each time period is the omitted year.
37
Table 8 – Branching and Profitability, 1926 - 1930
Dependent variable: Profit level of banks in different size categories in a state and by year.
Coefficient
Variable
Estimate
S.E.
Limited Branching
0.17 ***
( 0.02 )
Statewide Branching
-0.13 ***
( 0.04 )
No Law
-0.08 ***
( 0.03 )
Size Category 1
-1.34 ***
( 0.05 )
Size Category 2
-0.86 ***
( 0.04 )
Size Category 3
-0.47 ***
( 0.03 )
Size Category 4
-0.16 ***
( 0.03 )
Size Category 5
-0.14 ***
( 0.04 )
Size Category 7
0.21 ***
( 0.03 )
Size Category 8
0.28 ***
( 0.05 )
Size Category 9
0.35 ***
( 0.06 )
Size Category 10
0.65 ***
( 0.13 )
1926
-0.04
( 0.03 )
1927
-0.04
( 0.03 )
1929
-0.17 ***
( 0.03 )
1930
-0.96 ***
( 0.03 )
Intercept 7
-1.44 ***
( 0.03 )
Intercept 6
-0.47 ***
( 0.03 )
Intercept 5
0.57 ***
( 0.03 )
Intercept 4
1.47 ***
( 0.03 )
Intercept 3
2.20 ***
( 0.03 )
Intercept 2
3.25 ***
( 0.04 )
Observations (weighted)
37105
Log-Likelihood
-67719
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Estimated using an ordered logit. All information is from the Federal Reserve Board (1931).
The dependent variable is an ordered variable indicating the level of profitability weighted by the number
of banks that reported earning that level of profits. Weighted observations is the number of observations
multiplied by their respective weights. There are 8682 unweighted observations. The omitted branching
group is branches prohibited. Size category 6 and the year 1928 are also omitted.
38
Table 9 – Effects of Branching on National Bank Failures, 1926 - 1930
Dependent Variable: Number of failing banks in each group divided by the log of the number of banks in a group
(where a group is a set of banks in a similar size category in a state and by year)
Without Consolidation Index
Coefficient
Estimate
Intercept
No Law
-3.17 ***
S.E.
( 0.56 )
With Consolidation Index
Coefficient
Estimate
-3.26 ***
S.E.
( 0.56 )
0.07
( 0.30 )
0.09
( 0.30 )
Statewide Branching
-0.49 *
( 0.30 )
-0.41
( 0.30 )
Limited Branching
-1.04 ***
( 0.34 )
-1.14 ***
( 0.35 )
-1.16 **
( 0.57 )
Consolidation Index
Herfindahl Index
-2.65
( 2.02 )
-2.94
( 2.06 )
Business Fail Rate
-0.17
( 0.31 )
-0.17
( 0.31 )
Farm Price Index
-16.53 ***
( 3.72 )
-16.05 ***
( 3.73 )
Size Category 1
0.39
( 0.39 )
0.41
( 0.39 )
Size Category 2
0.57
( 0.36 )
0.57
( 0.36 )
Size Category 3
0.86 ***
( 0.33 )
0.86 ***
( 0.33 )
Size Category 4
0.25
( 0.37 )
0.26
( 0.37 )
Size Category 5
-0.65
( 0.48 )
-0.66
( 0.48 )
Size Category 7
-0.63
( 0.46 )
-0.63
( 0.46 )
Size Category 8
-2.12 *
( 1.08 )
-2.11 *
( 1.08 )
Size Category 9
-1.58 *
( 0.85 )
-1.62 *
( 0.86 )
Size Category 10
-21.79
( 99.0 )
-21.80
( 99.0 )
1927
-0.17
( 0.40 )
-0.16
( 0.40 )
1928
0.30
( 0.40 )
0.28
( 0.40 )
1929
0.72 *
( 0.38 )
0.76 **
( 0.38 )
1930
1.57 ***
( 0.33 )
1.65 ***
( 0.33 )
Caldwell Indicator
0.71 *
( 0.40 )
0.73 *
( 0.40 )
Observations
1688
1688
Log-Likelihood
-343.7
-341.8
Mean dependent variable: 0.065
Mean number of failures: 0.20
Mean (log banks in operation): 2.5
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Estimated using count data analysis. Information on the number of bank, size distribution
and on branching laws is from the Federal Reserve Board (1931). Other sources are: Population from the
1930 Census, business failure rate from the U.S. Department of Commerce, and farm prices and farm
foreclosures are from the Department of Agriculture (1939) and (1936) respectively. The omitted
branching group is branches prohibited. Size category 6 and the year 1926 are also omitted.
39
Table 10 – Effects of Branching on National Bank Failures, 1926 - 1930
Dependent Variable: Number of failing banks in each group divided by the log of the number of banks in a group
(where a group is a set of banks in a similar size category in a state and by year)
Without Consolidation Index
With Consolidation Index
Coefficient
Estimate
-3.00 ***
0.18
S.E.
( 0.59 )
( 0.30 )
Statewide Branching
-0.51 *
( 0.31 )
-0.43
( 0.32 )
Limited Branching
-1.01 ***
( 0.36 )
-1.09 ***
( 0.36 )
Ease-of-Merger Index
-0.65
( 0.56 )
-0.50
( 0.57 )
-1.00 *
( 0.59 )
Intercept
No Law
Consolidation Index
Coefficient
Estimate
-3.11 ***
0.19
S.E.
( 0.59 )
( 0.30 )
Herfindahl Index
-24.52
( 21.8 )
-28.17
( 22.3 )
Business Fail Rate
-0.17
( 0.32 )
-0.16
( 0.32 )
Farm Price Index
-15.75 ***
( 3.95 )
-15.19 ***
( 3.94 )
Size Category 1
0.37
( 0.40 )
0.39
( 0.40 )
Size Category 2
0.56
( 0.37 )
0.55
( 0.37 )
Size Category 3
0.90 ***
( 0.34 )
0.91 ***
( 0.34 )
Size Category 4
0.26
( 0.38 )
0.26
( 0.38 )
Size Category 5
-0.65
( 0.51 )
-0.66
( 0.51 )
Size Category 7
-0.57
( 0.48 )
-0.58
( 0.48 )
Size Category 8
-2.08 *
( 1.12 )
-2.07 *
( 1.12 )
Size Category 9
-1.69 *
( 0.93 )
-1.72 *
( 0.93 )
Size Category 10
-21.80
( 99.9 )
-21.83
( 99.9 )
1927
-0.10
( 0.40 )
-0.10
( 0.40 )
1928
0.30
( 0.42 )
0.26
( 0.42 )
1929
0.72 *
( 0.40 )
0.75 *
( 0.40 )
1930
1.66 ***
( 0.35 )
1.71 ***
( 0.35 )
Caldwell Indicator
0.66
( 0.41 )
0.66
( 0.41 )
Observations
2014
2014
Log-Likelihood
-441.9
-443.7
Mean dependent variable: 0.065
Mean number of failures: 0.20
Mean (log banks in operation): 2.5
Note. The symbols (***), (**), and (*) indicate statistical significance at the 1, 5, and 10 percent level,
respectively. Estimated using count data analysis. Information on the number of bank, size distribution
and on branching laws is from the Federal Reserve Board (1931). Other sources are: Population from the
1930 Census, business failure rate from the U.S. Department of Commerce, and farm prices and farm
foreclosures are from the Department of Agriculture (1939) and (1936) respectively. The omitted
branching group is branches prohibited. Size category 6 and the year 1926 are also omitted.
40
Appendix Table 1. Data Sources
Two main sets of data are employed. The first consists of state-level aggregate
data on national banks, covering the period 1921-1930. These data are used to compare
the prevalence of entry and exits between states with different branching regimes.
Information on number of banks, mergers, voluntary liquidations, and new banks are
compiled using the Annual Report of the Comptroller of the Currency (1921-1930).
The second contains additional information on national banks in each state
between 1926 and 1930. These data are drawn from the Federal Reserve Board of
Governors’ (1931) Report on Branch, Chain and Group Banking, Volume 9: Bank Profits
and further categorize banks by size (based on the sum of loans and investments). This
source also includes the information on bank profits by size category. Data from this Fed
report are also used for the construction of the Herfindahl index of banking concentration.
The four-firm bank concentration index is calculated using data on the four largest
national banks in each state from Polk’s Bank Directory (various years).
Bank failures for 1926-1930 are taken from the Annual Report of the Comptroller
of the Currency (1932, table 43, pp.208-28) and the Comptroller of the Currency’s
Statements of National Banks (1925-1929), and are matched to the appropriate size
category for the appropriate state using the information contained in Federal Reserve
Board of Governors (1931).
State bank data are matched to branching laws and to indicators of state economic
activity. Information on the branching laws for each state are from the Federal Reserve’s
Report on Branch, Chain and Group Banking, Volume 2: Branch Banking in the United
States. This reports the developments in each state’s branch banking laws from 1909 until
1931. Detailed information on state branching laws (including whether the de jure rather
than the de facto situation was used) is shown in Appendix Table 2. Business failures and
population estimates are from the U.S. Department of Commerce, Statistical Abstract of
the United States (various years). The index of lagged farm real estate farm prices are
from U.S. Department of Agriculture (1939) and farm foreclosure (bankruptcies) rates
are computed using data from U.S. Department of Agriculture (1936).
41
Appendix Table 2. Branch Banking Laws
State
Type
Alabama
Unit
Arizona
Statewide
Arkansas
Unit
California
Statewide
Colorado
Unit
Connecticut
Unit
Delaware
Statewide
Florida
Unit
Georgia
Multiple
Idaho
Unit
Illinois
Unit
Indiana
Unit
Iowa
Unit
Kansas
Kentucky
No Law
Notes
A few banks have branches
Branching had been practiced before law came into effect
State commissioner authorizes a few "exceptions"
Branching had been practiced before law came into effect
If charter allows
Branching allowed until 1927, banned until 1929 when it is permitted in Savannah
and Atlanta
Law against enacted in 1929
In 1902 the banking authority stated that law did not authorize, but "was not
construed as prohibitive." In 1909 the courts say the banks can't have branches but
Statewide* can have "offices to receive deposits and pay checks or transact other necessary
duties not requiring special discretion or business acumen." Observers at the time
noted little difference between these agencies and branches.
Louisiana
Limited
2 branches in the same parish
Maine
Limited
Only in the same county
Maryland
Statewide
Branching had been practiced before law came into effect
Massachusetts
Limited
Trusts can have branches in the same city, but prior to 1928 were limited to one
branch
Michigan
Limited*
No law - but a lot of branches in the city of the head office
Minnesota
Unit
Mississippi
Limited
Missouri
Unit
Montana
Unit
Nebraska
Unit
No law prior to 1927
42
Table 2, continued
Nevada
Unit
New Hampshire
No Law
New Jersey
Limited
New Mexico
Unit
New York
Limited
Commissioner was not aware of any law prohibiting branching
Allows "mercantile corporation which maintains a banking department to continue
operations at its branches." A clause in the law specifically for a particular
corporation.
Branching within cities of 50,000
North Carolina
Statewide
North Dakota
No Law
Though the law was construed as not permitting them
Ohio
Limited
In contiguous communities
Oklahoma
No Law
Oregon
Unit
Pennsylvania
Limited
Rhode Island
Statewide
South Carolina
Statewide*
South Dakota
No Law
Tennessee
Limited
Texas
Unit
Utah
Unit
Vermont
Virginia
Branching had been practiced before law came into effect
Branches are only permitted in places where national banks have branches already –
prior to 1927 had been in city of home office
No law per se, but instead what the capital requirements would be should the bank
have branches
Law somewhat unclear - branches permitted, but limited to the county of the banks’
home office.
Law changed in 1929. Prior to the change, branches were "not authorized." The
change allowed banks to establish "agencies" which were branches in all but name
In 1922 authorized for anywhere, in 1928 restricted to cities of 50,000. Considered
Statewide*
statewide because of the number of branches in place prior to the restriction.
Multiple
Washington
Unit
West Virginia
Unit*
Wisconsin
Unit
Technically allowed by law from 1925-1929 however the commissioner of banking
did not permit them
Law of 1921 says banks’ articles of incorporation (AoI) state the "place or places
where its offices be located" while the 1926 law has the AoI state the "place where
its office..."
Note: All types are as indicated by law (de jure) except were indicated by a * in which case the de facto type is used.
Please consult the table for the reason the de jure is not used.
Source. Federal Reserve Board of Governors (1931), Report of the Branch, Chain, and Group Banking Committee,
Volume 9.
Wyoming
Multiple
43